perf(store): reduce scoped-list allocations and skip unused artifact clones

[DeliciousBuding]

Summary

Closes #2337.

• Add a reproducible real Store.ListRuns / Store.ListArtifacts benchmark covering 100/1,000 resident rows and missing/sparse/dense/all scopes.
• Stage small filtered results locally and allocate their exact result size; overflow still allocates once for the dense case. Preserve one predicate invocation per row and insertion order.
• Clone artifact content-source metadata only after scope selection, while preserving detached returned values and non-nil empty results. Extend the existing query behavior/ownership tests.

Measured result and limits

At 1,000 resident rows, ten-run results allocate 1,792 B instead of about 172 KB. Ten-artifact results allocate 1,888 B / 11 allocations instead of 179,072 B / 1,001 allocations; missing artifact queries have zero allocations instead of 1,001. Dense/all result allocation is unchanged.

This is not a universal latency claim: after an initial +12.4% unfiltered-run timing, interleaved before/after controls showed overlapping ranges and +2.7%/+5.4% dense/all medians at 1,000 rows. The allocation improvement is retained with this explicit tradeoff; no production request distribution, RSS, retained heap or end-to-end throughput was measured.

• Baseline: 80 raw facade samples.
• 240 temporary strategy samples and rejected alternatives.
• 80 candidate facade samples, 48 interleaved controls, commands and acceptance evidence.

Validation

Passed on delivery files:

• Linux/WSL2 Go 1.26.5: CGO_ENABLED=1 go test -race -p 2 ./internal/store -count=1 -cpu=4 -timeout=10m.
• Windows: CGO_ENABLED=0 go test -p 2 ./internal/store -count=1 -cpu=4 -timeout=10m, go vet -p 2 ./internal/store, staticcheck ./internal/store (2026.1).
• Same real-facade fixture, five repetitions per case with allocation and ordered-ID assertions. Generic compiler escape diagnostics and git diff --check.
• Behavioral assertions cover one predicate visit, interleaved results around the small/dense transition, insertion order, non-nil empty slices, empty scope, nil content sources and mutation isolation. No timing thresholds or fixed-capacity tests.

All seven required checks passed for 5ab5fca23881e77031d093ea9bff5a35cc1202ec: validate, go-edge, go-hub, windows-go, windows-frontend, backend-required, frontend-required (CI run). The backend integration and fixture E2E lanes passed; frontend/native/real-E2E lanes were path-filtered or dispatch-only and skipped as reported by CI. CodeRabbit explicitly skipped review and is not counted as code review. L3 real-login/model and L4 packaged/release validation are not claimed.

Scope / hygiene

No API/schema/SQL/pool/index/debounce/runtime configuration change, new CI gate or production data. Existing architecture docs remain correct. Temporary comparison implementations were archived on the issue and removed from the worktree; only the benchmark, production helper and meaningful regressions remain. No runtime synchronization, rebuild, restart or deployment is included.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: e84a9694-8f4e-4cde-bc26-1d059e093659

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}